Companies spent the last two years lighting cash on fire to see which chatbot could write the best haiku about quarterly earnings. Now they want results. OpenAI and Anthropic built entire business models on the assumption that enterprise clients would keep paying for tokens like they were collecting Pokémon cards. Wrong.
The shift from "tokenmaxxing" to efficiency means corporations finally asked their AI teams what return they're getting on these seven-figure contracts. Turns out nobody knows. The CFOs who approved unlimited GPT-4 access are now demanding proof that the intern's automated email responder justified the budget line. It did not.
OpenAI raised billions betting that demand would scale forever. Anthropic positioned itself as the responsible alternative while charging the same prices. Both assumed enterprises would treat AI spending like cloud infrastructure, essential and infinite. Instead companies are treating it like that Peloton in the garage. Expensive. Underused. Mildly embarrassing.
The growth rates that convinced venture capital to dump money into these companies depended on customers never asking if the technology actually worked better than the previous version. They're asking now. Engineers are running cost-benefit analyses. Procurement teams are renegotiating contracts. Some companies are realizing they spent six months optimizing prompts when they could have hired one competent analyst.
This is what happens when hype cycles meet budget reviews. The AI labs promised exponential value. Customers expected it. Delivered value remained linear at best. Now the bills are coming due and nobody wants to explain to the board why the chatbot cost more than three full-time employees.
OpenAI and Anthropic will survive this. They'll adjust pricing. They'll pitch efficiency tools. They'll rebrand consumption as optimization. But the era of companies buying tokens just to say they're doing AI is over. Shame it took this long for anyone to check the receipts.
Photo by on Unsplash
Leave a Comment